1. Field of the Invention
This invention relates to a focus adjuster for a camera.
2. Related Background Art
In a known focus adjuster for a camera, a pair of foreground object images is formed by two light fluxes coupled from a foreground object through two different optical paths to an optical system. In the optical system a pair of photoelectric converting means converts the light fluxes into a pair of foreground object image signals consisting of a pair of discrete data, upon which predetermined correlation operations are performed while shifting the paired foreground object signals relative to each other, in order to determine a shift amount of the highest correlation degree and calculate a defocus of the optical system from that shift amount.
Such a focus adjuster will now be described with reference to FIGS. 20 and 21.
FIG. 20 is an example of application of the prior art focus adjuster to a replaceable lens type single lens reflex camera. Replaceable lens 100 can be removably mounted in camera body 105. With lens 100 mounted, a photographing light beam from a foreground object is passed through lens 103 to main mirror 107 provided in camera body 105. The light beam is partly reflected upwardly by main mirror 107 to a viewfinder (not shown). At the same time, the remaining part of the light beam is transmitted through main mirror 107 and reflected by sub-mirror 108 (as a focus detection light beam) to auto-focus module 120 (hereinafter AF module).
FIG. 21A shows an example of AF module 120. As is shown, the AF module comprises a focus detection optical system including field lens 122 and a pair of re-focusing lenses 124 and CCD (charge-coupled device) 125 having a pair of light-receiving sections A and B. In the above construction, light flux passing through paired areas symmetrical with respect to the axis of a light incidence pupil of lens 103 forms a primary image in the neighborhood of field lens 122 and is coupled through field lens 122 and re-focusing lenses 124 to form a pair of secondary images on a pair light-receiving sections A,B of CCD 125. When the primary image coincides with a film conjugate surface (not shown), the direction and relative position of the pair of secondary images with respect to the light-receiving sections of CCD 125 are predetermined by the construction of the focus detection optical system. More specifically, light-receiving sections A and B shown in FIG. 21B each consists of n light-receiving elements ai, bi (i=0 to n-1), and when the primary image coincides with the film conjugate surface, a foreground object image is formed in substantially equal areas on light-receiving sections A and B. When the primary image is formed on a surface shifted from the film conjugate surface, the relative position of the pair of secondary images on CCD 125 is shifted from the equal area position noted above according to the direction of shift of the axial direction of the primary image (i.e., whether the shift is pre- or post-focus. In the case of the post-focus, for instance, the positional relation between the pair of secondary images is relatively spread, and in the case of the pre-focus it is narrowed.
Light receiving elements ai and bi forming light-receiving sections A and B consist of charge storage elements such as photo-diodes and can accumulate charge for a charge accumulation time corresponding to the illumination intensity on CCD 125 to render the light-receiving element output to be of a level suited for a subsequent process to be described later.
Returning to FIG. 20, interface 112 and memory 113 control the start and end of charge accumulation in CCD 125 by providing a control signal to CCD 125 according to charge accumulation start and end commands from calculation and control unit (AFCPU) 110 A light-receiving element output signal is sequentially transferred to the interface according to transfer clock signals supplied to CCD 125. A/D converter means provided in the interface samples and A/D converts the light-receiving element output, and A/D converted data (2n pieces) corresponding to the number of light-receiving elements are stored in memory 113. AFCPU 110 performs well-known operations for focus detection according to the stored data to determine a defocus extent corresponding to the difference between the primary image and film conjugate surface.
AFCPU 110 controls the form of display on AF display means 114 according to the result of the focus detection calculation. For example, AFCPU 114 provides a control signal such that a triangle display section becomes active in case of pre- or post-focus and a round display section becomes active in case of in-focus. AFCPU 110 also controls the direction and extent of driving of AF motor 109 according to the result of focus detection operation, thus moving lens 103 to an in-focus position. According to the sign of the defocus extent (whether pre- or post-focus), AFCPU 110 generates a drive signal to cause rotation of AF motor 109 in a direction such that focusing lens 103 approaches the in-focus point. The rotational motion of the AF motor is transmitted by a camera body transmission system consisting of gears provided in camera body 105 to camera body coupling 109a provided in a mount section of camera body 105, on which lens 100 is mounted. The rotational motion transmitted to camera body coupling 109a is further transmitted through lens transmission system 102 including lens side coupling 101 (fitted with coupling 109a) and gears provided in lens 100, thus causing movement of lens 103 toward the in-focus position.
Lens 100 includes lens inner calculation and control means (lens CPU) 104, and it supplies necessary AF data such as the number of rotations of coupling 101 per unit displacement of lens 103 to AFCPU 110 through a lens contact provided in the mount section and camera body contact 106.
The focus control operation in AFCPU 110 will now be described in detail. All image outputs obtained from image sensors A and B in FIG. 21B are subjected to an image output shift as shown in FIGS. 22A and 22B to calculate the image shift extent. More particularly, denoting the image outputs of image sensors A and B by a0 to a39 and b0 to b39 (here n=40), FIG. 22A shows a case when shift L is L=20, FIG. 22B shows a case when L=0, and FIG. 22C shows a case when L =-20. At each shifted position, the correlation of corresponding images of image sensors A and B is obtained to be compared to the extent of correlation concerning each shift position. A shift extent of the best correlation is determined to be the image shift extent. Lens 103 and other parts are driven according to a focus control operation such as correlation calculation, to determine this image shift extent.
FIG. 22E shows a different way of expressing the method of determining such a shift. In the matrix of the Figure, points shown by a dot mark represent comparative picture elements. In this case, a shift number (i.e., shift extent) range of -20 to 20 constitutes the subject of calculation. The correlation calculation time is proportional to the number of blocks contained in this range. Problems in this calculation will now be discussed with reference to FIGS. 15 and 16.
FIG. 15 shows viewfinder screen 3100 of a single-lens reflex camera and its focus detection field. In the prior art, focus detection zone frame 3101 is so narrow that a depth to be described later does not enter the zone, and it is used for focus detection. In the in-focus state, image sensors A and B are located at positions equivalent to the narrow frame shown by the solid line in FIG. 15. Reference numeral 3102 designates marks provided in the viewfinder field in correspondence to a boundary in a detectable range. However, where focus detection can be effected in only a narrow focus detection area 3101, although there is no problem if there is an adequate pattern of a foreground object in focus detection zone 3101 as shown in Fig. 16A (FIGS. 16A and 16C showing the focus detection zone at the center of FIG. 15 on an enlarged scale, and FIGS. 16B and 16D showing output ai of image sensor A), if the pattern vanishes due to movement of the camera as shown in FIG. 16C, H) disability of focus detection results. In such a case, the photographing lens unnecessarily starts scanning, which is cumbersome.
Accordingly, in order to maintain and improve the accuracy of the correlation calculation noted above and also broaden the image area in the viewfinder field that is capable of focus detection, it is necessary to increase the number of picture elements (number of light-receiving elements), i.e., number of blocks in FIG. 22E, to thereby increase the focus detection area of the focus adjuster. However, increasing the number of picture elements to meet this requirement presents various problems as follows.
(1) With a wide focus detection area, there is a high possibility that foreground objects at different distances or depths are found in the area. Therefore, it is necessary to divide the wide focus detection area into a plurality of local focus detection areas and perform a focus control operation for each local focus detection area. In addition, there are problems concerning the division of the focus detection area.
(2) There are problems in the selection of defocus from among those of the local focus detection areas for indication and driving and in the manner in which the user's intention is to be reflected.
(3) There are problems accompanying the wideness of the focus detection area such as a problem of selecting a local focus detection area, with respect to which AGC (to be described later) is to be provided.
(4) Although increasing the image area capable of use for focus detection has a merit of increasing the shift (to be described later) during image correlation to permit an increased defocus area to be provided for focus detection, the prior art method of calculation is time-consuming.
Problems in (1) will be discussed.
First, the problems presented by the division of the focus detection area in the prior art method will be discussed.
As for one of image sensors A and B shown in FIG. 26, namely image sensor A, there are two well-known methods of division, i.e., a first method shown in (.alpha.), in which local focus detection areas R1 to R3 free from overlap are defined, and a second method in (.beta.), in which partly overlapped local focus detection areas R1' to R3' are defined. An example of focus control calculation is disclosed in U.S. Pat. No. 4,812,869. In this example, the method of image shift control calculation is changed when the defocus is large and when it is small. More specifically, when the defocus is large, the image shift is calculated by shifting all image outputs of image sensors A and B shown in FIG. 21B in a manner as shown in FIGS. 22A to 22C. When the defocus is small, on the other hand, image sensor A is divided to define overlapped local focus detection areas R1' to R3' as shown in FIG. 22D for detecting correlation between image sensors A and B.
Now, the two methods of division noted above will compared with respect to their merits and demerits. A case will be considered, in which a foreground object having a light intensity distribution as shown in FIG. 27A is projected on image sensor A. In this case, the image shift can be calculated by either of the methods (.alpha.) and (.beta.) in FIG. 26, with local focus detection area R2 and image sensor B in the former method and with local focus detection area R2' and image sensor B in the latter method. (The image shift can be detected when and only when there is a change in the light intensity distribution of the object.)
Now, a case will be considered, in which a foreground object light intensity distribution having an edge as shown in FIG. 27B is projected on image sensor A. In the division method (.alpha.), the edge between bright and dark portions is just on the boundary between local focus detection areas R1 and R2. In this case of presence of a portion of large brightness change (i.e., large data amount) at an end of a focus detection area, the large brightness change portion is shifted out of and into the area when the shift of the optimum correlation is obtained by shifting the image signal projected on the other image sensor. This reduces the accuracy of image shift calculation. In the division method (.alpha.), this applies to both detection areas R1 and R2 when there is an edge at the borderline between areas. That is, it is only when the edge is at this position that the accuracy of detection deteriorates, or disability of detection results.
With the division method (.beta.), on the other hand, such an edge is completely contained at least in either area R1' or R2', and thus the above problem is not presented.
Now, a case will be considered, in which there is a depth or distance in an image. In such a case, an image as shown in FIG. 28A is projected on image sensor A, while an image as shown in FIG. 28D is projected on image sensor B. In this case, the close image areas and also the distant image areas can be overlapped by shifting areas, but the borderline areas can not be overlapped by shifting. Therefore, an area where close and distant images coexist can not be detected. First, when the image shown in FIG. 28A is projected on image sensor A, in the case of the division method (.alpha.) the borderline between the close and distant images is found substantially at the borderline between areas R1 and R2. Thus, a close image can be detected using area R1, while a distant image can be detected using areas R2 and R3. In the case of the division method (.beta.), images with depths are contained in both areas R1' and R2'. Therefore, detection is impossible in these area. Table 1A shows what is described above. FIGS. 28B and 28C show cases where the image on image sensor A is located at slightly shifted positions. Tables 1B and 1C show the possibility of detection in these cases.
TABLE 1 ______________________________________ R1 R2 R3 R1' R2' R3' ______________________________________ A close- distant distant undetec- undetec- distant range view view table table view view B close- undetec- distant undetec- undetec- distant range table view table table view view C close- undetec- distant close- undetec- distant range table view range table view view view ______________________________________
It will be seen that a case where there is a depth of image, the division method (.alpha.) of defining local focus detection areas without overlap has a higher possibility of detection. This is obvious from the fact that a depth found in overlap portions of local focus detection areas disables detection in both the areas. As is shown, the division methods (.alpha.) and (.beta.) have their own merits and demerits.
Now, the problem in (2) concerning the selection of defocus among those of the local focus detection areas for indication and driving, will be discussed. When distant, intermediate and close range views of images are projected on image sensor A divided into local focus detection areas R1 to R7 as shown in FIG. 14, the photographer should determine the image to be focused. If the camera itself determines the view for focusing, it may sometimes go counter to the photographer's will. To be able to accurately focus an intended foreground it is necessary to set in advance a narrow focus detection area such that a depth will not enter the area.
In relation to this, Japanese Patent Laid-Open No. Sho 63-11906 proposes a system, in which focusing is effected according to the result of detection using three detection areas 3601 to 3603 provided in a viewfinder screen as shown in FIG. 19. In this case, there is a case when what can not be detected in central area 3601 can be detected in end area 3603. In such case, since the end area is spaced apart from the central area, there is a high possibility that a distant thing (for instance a tree outside a window) different from a thing in the central area (for instance a room wall) is focused.
The problem in (3) will now be described.
When continuous image sensor A (such as that with local focus detection areas R1 to R7 shown in FIG. 14) is used for detecting a plurality of foreground object images in a wide area, the brightnesses of the individual foreground objects may vary from one another extremely. In such a case, the way of providing AGC is important. Problems in this connection will be discussed.
First, a case is considered, in which there is A/D conversion capacity of about 8 bits for A/D converting analog image output into digital data to be stored in memory 113. With this order of dynamic range, if the brightness varies extremely, the output level can be optimized only for a portion of foreground objects through storage time control (i.e., through AGC). A case will now be considered, in which a wall with a poster applied thereto and a nearby window are projected on an image sensor for detecting a viewfinder screen central portion as shown in FIG. 17A. FIG. 17B shows a case when AGC is provided such that an image output in a predetermined central area is a predetermined peak (for instance 150), and FIG. 17C shows a case when AGC is provided such that the entire image output has a predetermined peak (for instance 150).
It will be seen that in the case of FIG. 17C the central image brightness is too low to be detected by corresponding local focus detection area R4 of the image sensor. Accordingly, the local focus detection area H) of image sensor A is shifted outwardly to areas R3, R5, R2 and R6, and focus detection first becomes possible in area R6. However, the photographer is thinking that naturally the poster at the center is focused. Therefore, there is a possibility of going counter to the photographer's will.
The problem in (4) will now be discussed.
As noted before in connection with the prior art, in the calculation of correlation between image sensors A and B the calculation time is substantially proportional to the number of blocks in FIG. 22E. In the example of FIG. 22E, the number of picture elements is 40 pairs. However, if this method is used where the number of picture elements is 80 pairs, the calculation time is increased to several times.
According to U.S. Pat. No. 4,636,624, the image on image sensor A is divided into three blocks as shown in FIG. 25, and when comparing image sensors A and B in an area of small defocus and small shift number L as shown in FIG. 23, correlation C(L) between the picture element row of the second block shown at .beta. and data of image sensor B after shift is obtained, while for areas .alpha. and .gamma. the correlation C(L) is calculated for the first and third blocks with respect to data of image sensor B after shift.
Here, the correlation C(L) is the value given as EQU C(L)=.THETA..vertline.a.sub.i -b.sub.i+L.vertline.
In this method, as shown in FIG. 23, the correlation is discontinuous at shift numbers corresponding to the borderlines between adjacent blocks